Is need text in training corpus or only wav files?

Thanks you!

[seungwonpark]

Both text and audio are required.

Is can mix languages? or English only is supported?

[seungwonpark]

Cross-lingual conversion will be possible if it's trained with a dataset of multiple languages. This repo does not directly support such training strategy, but you'll be able to try that by changing some parts of the code!

I have five chinese speakers, make training is OK. How long each voice should have? How many speaker should have?